GridConfig class to help with instantiation

[jjuyeonkim]

Description
This originally came up out of PR #246, which is related to Issue #64.

While trying to clean up the soon-to-be-removed NDSL Namelist, we needed a way to extract grid parameters from a namelist file. A simple dict was used to move forward, but it really makes sense to have a class replace that.

Potential solution
Currently, in the translate tests (conftest.py), we do something like the following to create a dict to eventually use to initialize the grid:
grid_params = grid_params_from_f90nml(namelist)

Rather than return a dict, a potential class could be (suggestion from @romanc and @FlorianDeconinck):

class CartesianGridConfig
    nx: int,
    ny: int,
    nz: int,
    
class CubeSphereGridConfig:
   layout: tuple[int, int]
   cartesian_dims: CartesianGridConfig

[fmalatino]

I like this idea. Would this also make use of dacite or something similar to generate this (data)class?

[jjuyeonkim]

It could use dacite. That's a possibility, especially since we're already using it elsewhere. For NDSL Issue #64 , I'm planning on making a helper function in pyfv3 and pyshield to go straight from a dict created from the namelist file to their respective DynamicalCoreConfig and PhysicsConfig objects by way of dacite. If that ends up being okay, then maybe this can be similar... but maybe *GridConfig instead of *GridParams? This is just an idea though.

[FlorianDeconinck]

Yes GridConfig, we should keep continuity of the naming scheme

[jjuyeonkim]

Adding a note -- I just realized that there are GridConfig classes defined in pace's grid.py.

The SerialboxGridConfig seems analogous to the grid_params dict used in PR #246. I didn't realize that at the time. There's also a GeneratedGridConfig and ExternalNetcdfGridConfig. It seems like a generalized version could be added to NDSL and then used within pace.